A teammate suggests automating the entire test suite to eliminate manual testing. How would you respond?
technical-conceptual · Junior level · software-engineering
What the interviewer is really asking
Assesses whether the candidate understands the genuine limits of automation and the irreplaceable role of human exploratory and usability testing, while engaging the idea constructively.
What to say
- Acknowledge the upside — full automation of stable regression checks is a real win for speed and confidence — so you're engaging the idea, not dismissing it.
- Explain what automation can't replace: exploratory testing that uncovers the unexpected, usability and visual judgment, and assessing new or ambiguous features that have no fixed expected output to assert against.
- Reframe the goal as the right balance, not 100% automation — automate the repetitive deterministic checks and keep deliberate human testing where judgment and curiosity find the bugs scripts miss.
What to avoid
- Flatly agree and treat manual testing as legacy work to phase out — that loses the bugs only humans catch.
- Shut the teammate down without acknowledging the real benefits of automation, coming across as resistant rather than thoughtful.
- Pretend automation can do exploratory or usability testing, which misunderstands what those activities are.
Example answers
Strong: I'd say yes to automating the stable, repetitive parts — regression and smoke tests should absolutely run themselves so we move faster. But I'd push back gently on 'eliminate manual testing', because automation only checks the cases we already thought of. Exploratory testing is where a person pokes at the new feature and finds the bug nobody scripted, and usability — does this flow actually make sense — isn't something a script can judge. So I'd reframe it as automating the deterministic regression layer and keeping a focused human exploratory pass, rather than aiming for zero manual testing.
Weak: If they think we can automate everything, I'd go along with it — automated tests are more reliable than people anyway, and it'd free up time. We'd just have to write enough tests to cover all the cases.